☰

AstraZeneca

Twenty new medicines by 2030

Industry
Biopharmaceuticals
Headquarters
Cambridge, England, United Kingdom
Public information as of
January 2026

A4BEE prepared this analysis from publicly available sources. It reflects our own reading of AstraZeneca's published strategy and is not endorsed by, or produced in cooperation with, AstraZeneca. Company website

Strategic priorities

AstraZeneca has set a public goal of reaching $80 billion in total revenue by 2030 with twenty new medicines targeted to launch, and the Polish operation sits inside the delivery plan. Poland hosts one of six AstraZeneca Global Clinical Trials Centers in Warsaw, a ten-times-larger Krakow R&D data capability centre, and a Global Finance Services centre that acts as the financial engine for 85 countries. The common thread across these is the same set of data: clinical, laboratory and operational.

AstraZeneca has committed to a 98 percent reduction in Scope 1 and 2 emissions by 2026 and full carbon neutrality by 2030 under the Ambition Zero Carbon programme, with the Warsaw site already running on 100 percent renewable electricity. The next scale of the work is Scope 3, which means tracking the energy and product footprint of more than a thousand suppliers and reporting it as a continuous number rather than as an annual filing.

Recent acquisitions including Alexion, Gracell Biotechnologies and Fusion Pharmaceuticals added new modalities (rare disease, cell therapy and radioconjugates) and new data shapes. Integrating these alongside the company's existing platforms is the operational form the $80 billion goal takes inside IT, where the IT-2025 programme is the umbrella under which AI and modernisation work runs.

The clinical-trial cost per drug asset has been reported at $2.2 billion, with clinical trials consuming around 70 percent of that figure, so the time saved or lost on trial startup and execution is a direct line on the financial plan. Sites like Warsaw and Krakow are the practical entry points for the modernisation work because the trial data pipeline already lives there.

Challenges we see

  • Operations R&D

    Consolidating clinical trial data across systems and countries

    The Warsaw Global Clinical Trials Center manages the data pipeline for AstraZeneca's entire portfolio across six therapeutic areas. Clinical trials account for around 70 percent of the $2.2 billion average cost per drug asset, and the same pipeline is the entry point for AI work in R&D.

    Where clinical, EDC, RTSM and operational data live in separate systems, the same trial is reviewed from four different perspectives and AI models cannot see it as one body of evidence. Bringing those views onto a single time-aligned spine is what makes the AI investment pay back.

  • Digital Integration

    Moving AI pilots into production-ready R&D use cases

    AstraZeneca's leadership has publicly committed to AI-driven R&D, and the Global Chief Digital Officer describes AI as fundamental to how the company develops drugs. Across pharma, around 11 percent of AI implementations are reported as delivering benefit and 89 percent of AI projects do not reach production.

    The work is less about new models than about the data quality, governance and validation evidence that lets an AI system operate inside a GxP-controlled environment. Naming the gap is one thing; building the audit chain that lets regulators and operators trust the answer is the harder and longer piece.

  • Digital Integration

    Connecting clinical, IT, QA and finance data sources

    The Warsaw and Krakow hubs run Clinical Operations, IT, QA and Finance on separate data silos. EDC (Electronic Data Capture) and RTSM (Randomization and Trial Supply Management) integration is reported as a persistent challenge, and several laboratory workloads still rely on standalone Industrial PCs (IPCs).

    Siloed systems force every cross-functional question to be reconciled by hand and turn a single hardware failure into a stopped trial. An architecture that publishes data once and consumes it across functions, and infrastructure that survives a single node failure, changes what each of those questions costs.

  • Compliance Regulatory

    Producing regulatory and inspection evidence as data flows

    The EU AI Act and FDA GxP frameworks together raise the documentation bar for AI used in clinical trials, and the Polish operation serves global regulatory submissions. The integration of acquired companies (Alexion, Gracell, Fusion) adds new documentation shapes to an already broad submission load.

    Inspections, deviation reports and change-control evidence scale with the number of studies and sites, so the volume itself becomes the bottleneck. Generating that evidence from the same systems that run the work, rather than assembling it after the fact, is the change that keeps the queue moving.

  • ESG Sustainability

    Reporting Scope 3 progress across a thousand suppliers

    AstraZeneca's Ambition Zero Carbon targets a 98 percent cut in absolute Scope 1 and 2 emissions by 2026 and full carbon neutrality by 2030. The Warsaw site is already on 100 percent renewable electricity; the harder work sits in Scope 3 across a value chain of more than 1,000 suppliers.

    Where Scope 3 is reconciled from supplier self-declarations at year-end, the report is a snapshot rather than a measurement. Continuous data exchange with the supplier base moves Scope 3 from a disclosure exercise into a management metric that the operations team can actually move.

Opportunities, by urgency and business impact

Each bubble is one opportunity, numbered to match the list below. Further right means it bites sooner; higher means a bigger effect on the business. A bigger bubble means a bigger implementation effort.

Source: A4BEE analysis of public sources

  1. Consolidating clinical, EDC and RTSM data onto one spine

    Clinical trial data is reported to be spread across four systems, six countries and seventeen formats, which prevents AI modelling on the underlying body of evidence and delays the daily picture of trial status that programme teams need.

    A unified data platform that defines clinical, EDC, RTSM and operational entities once and loads all sources against that model gives every consumer, from trial managers to AI pipelines, the same time-aligned view.

    • AstraZeneca Annual Report and Form 20-F 2024
    • BPCC, Accelerating life-saving innovations, January 2026
    • Veeva Program Manager, Global Clinical Solutions job description, AstraZeneca Careers

  2. Replacing manual KPI calculations with platform-published metrics

    Global Finance Services in Warsaw and laboratory operations rely on manually tracked KPIs assembled in spreadsheets. Critical biological metrics such as Viable Cell Density (VCD) are tracked separately from live process trends, which lengthens the time between an event and its visibility to a manager.

    Move the calculation of these KPIs into the data platform itself, publish them as data services, and let dashboards and operational reports consume the same numbers rather than each function recomputing its own version.

    • Global Finance Services, AstraZeneca Careers
    • Senior Clinical Finance Business Partner, AstraZeneca Careers, Warsaw

  3. Migrating standalone laboratory IPCs to a clustered architecture

    Several laboratory workloads still run on standalone Industrial PCs (IPCs) that act as single points of failure: a hardware failure during an experiment halts data collection, and the workstation itself becomes a constraint on what the lab can change.

    Migrate those workloads to a container-based cluster architecture so the software can move between nodes, the data collection path has redundancy, and a single hardware event does not stop a study.

    • AstraZeneca Krakow job postings, AstraZeneca Careers
    • Biologics Digital Transformation Lead, AstraZeneca Careers, Södertälje

  4. Bringing computer-vision monitoring to bioreactor operations

    Cell therapy manufacturing acquired with Gracell and radioconjugate work from Fusion require precise controls, and the manual monitoring that has worked for traditional biologics is a slow reaction path in modalities where a deviation compounds quickly.

    Apply external-camera computer-vision monitoring and adaptive dosing control to bioreactor operations, so critical bioprocess parameters are observed continuously and adjustments are made on the same timescale as the process itself.

    • Gracell Biotechnologies acquisition disclosure, AstraZeneca Annual Report 2024
    • Fusion Pharmaceuticals acquisition disclosure, AstraZeneca Annual Report 2024

  5. Producing Scope 3 supplier sustainability evidence as continuous data

    Scope 3 reporting under Ambition Zero Carbon covers more than 1,000 suppliers, and the existing data exchange is largely manual and retrospective. Each reporting cycle starts from a fresh request for self-declared values rather than a running measurement.

    Replace the annual request cycle with continuous data exchange so supplier energy use, product footprint and emissions move into a shared model, and the Scope 3 number is monitored through the year instead of reconstructed at year end.

    • Ambition Zero Carbon, AstraZeneca Sustainability
    • Sustainability Data Annex 2024, AstraZeneca

What we'd propose

  • Enterprise AI

    Unified data platform for the Global Clinical Trials Center

    An ontology-based data platform that defines clinical, EDC, RTSM and operational entities once, then loads all AstraZeneca sources against that model so trial managers, AI pipelines and operational dashboards read the same time-aligned evidence.

    • Shared clinical and operational ontology

      One agreed set of terms

      Define clinical visit, subject, sample, RTSM arm, EDC record and operational event as explicit entities with agreed relationships, so a query written once returns comparable answers across all source systems instead of several dialects of the same table.

    • Pipelines from EDC, RTSM and operational systems

      Loading the trial pipeline

      Build ingestion for the Warsaw clinical stack, including EDC, RTSM and adjacent operational sources, with schema validation at the boundary so a record that does not match the ontology fails loudly rather than silently.

    • Retrieval and analytics on the combined model

      Questions answered without new extracts

      Expose the model through dashboards and a retrieval layer so programme teams, medical leads and the AI work in IT-2025 can query the trial estate directly, rather than commissioning a new extract for each question.

    • Every consumer reads the same time-aligned trial picture instead of reconciling four versions of it.
    • AI models in IT-2025 train and run on a unified evidence base rather than on reconciled extracts.
    • Programme teams get a daily view of trial status, so the time between an event and its visibility collapses from days to minutes.

    We've done this before

    See how we do Enterprise AI
  • Enterprise AI

    Platform-published KPIs for finance and laboratory operations

    An ontology-based data platform that publishes critical KPIs such as cost-of-service, productivity and Viable Cell Density (VCD) as platform-computed data services, removing the Excel layer between the source system and the report.

    • Semantic mapping of financial and operational KPIs

      KPIs as data, not spreadsheets

      Map cost-of-service, productivity, VCD and other recurring KPIs to universal business classes in the ontology, so the same metric has the same meaning in finance, operations and the laboratory.

    • Automated KPI engineering

      Calculations done by the platform

      Compute the KPIs inside the data platform from validated source records, with a documented calculation path that finance, operations and the lab can all reference.

    • Self-service consumption of platform metrics

      Same number, every dashboard

      Expose the KPIs through dashboards and a retrieval layer so each function reads the platform-published value rather than recomputing its own, and a metric question is answered from one place.

    • Report generation cycles shorten because the numbers come from the platform rather than from a manually refreshed workbook.
    • Finance, operations and the lab look at the same number, so reconciliation stops being a recurring meeting.
    • The KPI definition becomes a reusable artefact, so the next metric is added against the same model.

    We've done this before

    See how we do Enterprise AI
  • Digital CDMO

    Containerised cluster architecture for laboratory workloads

    A migration from standalone Industrial PCs (IPCs) to a container-based cluster architecture for laboratory workloads, so the software is decoupled from the machine it runs on and a single hardware event does not stop a study.

    • Cluster architecture design

      Workloads that move between nodes

      Design a clustered environment where laboratory workloads can move between nodes without losing state, and document the trade-offs so the choice between N+1 redundancy and active-active is made deliberately.

    • Containerised orchestration of lab applications

      Software decoupled from hardware

      Encapsulate laboratory applications in containers so deployment, rollback and recovery happen against an image rather than against a specific machine, and the same image runs at any site.

    • Backup, recovery and operational runbook

      Recovery that is rehearsed

      Define backup, disaster recovery and operational runbooks against the new architecture, and rehearse the recovery path so the first time it runs is not during a live failure.

    • A single hardware failure does not halt data collection during an experiment.
    • A laboratory workload is reproducible across sites, so the Krakow and Warsaw hubs run the same validated image.
    • Recovery becomes a documented operation rather than an emergency.

    We've done this before

    See how we do Digital CDMO
  • Digital CDMO

    Computer-vision monitoring for bioreactor operations

    An external-camera computer-vision system that observes bioreactor conditions continuously and drives adaptive dosing for parameters such as foam level and antifoam addition, so the cell therapy and radioconjugate processes run with the same oversight as a traditional biologics line.

    • Non-invasive visual monitoring

      Cameras, not probes in the product

      Deploy external cameras and image processing to detect foam level, color changes and other surface signals without physical contact with the biological product.

    • Adaptive dosing control

      Multiple dosing strategies

      Implement PWM (pulse-width modulation), vector and interval dosing modes so the controller can match the foam behaviour, including slow-rising and flash foam patterns.

    • 24/7 autonomous operation with operator override

      Continuous monitoring, named operator

      Run the controller continuously with a defined operator override, so the response time to a deviation is seconds rather than the time it takes for an operator to look at a screen.

    • Cell therapy and radioconjugate processes are observed at the same rate as they evolve, so a deviation is acted on within seconds.
    • The control loop is the same at every site, so a process behaves consistently across the global manufacturing network.
    • Operators stay in the loop as named approvers rather than as the only line of detection.

    We've done this before

    See how we do Digital CDMO
  • Agents

    AI agents for clinical, regulatory and supplier documentation

    Narrow, reviewable agents that take the repetitive part of clinical and regulatory documentation work: drafting deviation and change-control summaries from source records, checking a submission against its template before review, finding every controlled document a standards change touches, and assembling Scope 3 supplier evidence from continuous data exchange. A named person approves every output.

    • Drafting from source records

      First drafts from system data

      Generate the first draft of a deviation, change-control, periodic-review or trial-startup document directly from the underlying system records, so the author edits and judges rather than assembles from scratch.

    • Template and completeness checking

      Gaps found before review

      Check a submitted document against its template and the site's own checklist, returning missing or inconsistent sections before the document enters the human review queue.

    • Change impact search across the document set

      Which documents a change touches

      When a standard, method or regulation changes, retrieve every controlled document that references it and rank them by how directly they are affected, so the update scope is known on day one rather than discovered in audit.

    • Review queues move faster because documents arrive complete and the first draft is already sourced.
    • The scope of a standards or regulation change is established by search rather than by recollection.
    • Every output is traceable to the records it came from and signed off by a named reviewer.

    We've done this before

    See how we do Agents

Digital maturity: today and target

Scored out of 100 across six dimensions. The target is what AstraZeneca's own published ambition implies — not a perfect score.

Source: A4BEE analysis of public sources
Data integration 35 → 82
Clinical, EDC, RTSM and operational data are reported to live in separate systems across multiple countries and the Polish hub. The unification work is ahead of the company rather than behind it, and the target reflects a single time-aligned spine rather than a portfolio of reconciled extracts.
AI in regulated R&D 28 → 75
AstraZeneca has publicly committed to AI-driven R&D, but across pharma only around 11 percent of AI implementations are reported as delivering benefit. The current state reflects leadership vision rather than production-ready systems in GxP environments.
Infrastructure resilience 42 → 85
Several laboratory workloads still run on standalone Industrial PCs that act as single points of failure. A clustered container architecture is well understood in the wider industry and the work is moving the Krakow and Warsaw hubs onto it.
Process automation 45 → 82
Critical KPIs are tracked in manually maintained spreadsheets and biological metrics such as VCD are tracked separately from live process trends. The automation target reflects platform-published KPIs rather than each function recomputing its own.
Operator UX 50 → 80
UX is named as a recurring friction in the Polish operation, with alert fatigue and dashboard misalignment reported. The target reflects dashboards that mirror the physical lab layout and reduce cognitive load rather than a separate change-management programme.
Scope 3 reporting 55 → 88
The Warsaw site is already on 100 percent renewable electricity, so Scope 1 and 2 are largely solved at the Polish hub. Scope 3 across more than 1,000 suppliers is largely retrospective and the target reflects continuous data exchange rather than annual reconciliation.

Check this yourself

Our Service Portal has free self-assessments and market comparisons. These are the ones that line up with what we've read above — no sales call required.

Think we've read this right?

Talk to us

Related reading

This is an independent analysis prepared by A4BEE from publicly available information as of January 2026. It reflects A4BEE's own interpretation and opinion, is not affiliated with, endorsed by, or verified with AstraZeneca, and may be incomplete or inaccurate. All company names and trademarks are the property of their respective owners. To request a correction or removal, contact [email protected].